It appears that you're asking when voicing ceases during the closure period of "g" in "bag", where the closure is the selected piece. Since you don't have physiological data on vocal fold vibration, you would estimate it from the acoustic waveform, looking for quasi-periodic variations in the waveform. Notice to the right that the waveform is fairly flat, but not totally flat, because there's usually some noise.
At the left edge of the closure for about 10 msc there is relatively clear voicing. There is still a low-amplitude of peaks and valleys for most of the closure, using the section after the release as your point of reference. If you could compare "bag" and "back", you could more clearly see whether this is present in /k/ as well. If it is, you should disregard that variation as not reflecting "voicing", since /k/ isn't voiced. Otherwise, if you have low amplitude quasi-periodic variation during consonant closure, then I would say that it counts as "voicing" -- unless you've been told otherwise by your instructor. It is difficult to locate the end of that amplitude variation by visual inspection, and does require a measure of background noise. (60 cycle AC noise would persist through the recording and not just be present during production of certain phonemes).